AI API price comparison 2026: DeepSeek V4 Flash — .014/1M input GPT-4.1-mini — .40/1M Gemini 2.5 Flash — .15/1M Claude Haiku 4.5 — .80/1M That is 28x cheaper. A
New benchmarks show Google's Gemini-3.5-flash matching OpenAI's GPT-5.5 on long-form Russian content at a 2.5x lower cost. Chinese models are also demonstrating significant price-performance advantages, with DeepSeek V4 Pro becoming more cost-effective after a price reduction. Tencent's Hy3-preview offers a particularly low-cost option for A-tier performance. Several sources highlight DeepSeek V4 Flash as being up to 28 times cheaper than GPT-4.1-mini, accessible via Token China with an OpenAI-compatible API. AI
IMPACT Accelerates adoption of cost-effective LLMs, pressuring incumbents like OpenAI and Anthropic to compete on price.